One of my instructors way back at Cornell was a grad student named Kevin Walsh, now an associate professor at Holy Cross, who had solved this problem in a bit of a radical way. Context was that back then you had Gnutella, LimeWire etc.—peer-to-peer types of networks—and you wanted to be able to enable people to rank media a certain number of "stars" for its quality because otherwise people post not-safe-for-life content with nice endearing names and others mass-download that content and then get grossed out. But you publish this and then people immediately use the anonymity afforded by the Internet to spam the upvote button as they distribute their garbage.

My understanding from Walsh was that the problem was essentially an economics problem—you want to incentivize good behavior and deincentivize bad behavior—and once you understood this you could use the network to correct itself, essentially saying “if you use your upvotes like the typical user does then the typical user will trust you, if you use your upvotes like the typical spammer does then you’ll instead end up in a clique with typical spammers.”

Some googling reveals that the page is still alive on cornell.edu [1].

[1] http://www.cs.cornell.edu/people/egs/credence/